This repository provides code for fine-tuning and post-training BERT models on review-based NLP tasks including review reading comprehension (RRC), aspect extraction (AE), and aspect sentiment classification (ASC). It implements techniques from multiple research papers (NAACL 2019, COLING 2020, EMNLP 2020) that explore domain-specific BERT adaptation and pre-training strategies. The project includes preprocessing tools for ABSA datasets and supports cross-domain model deployment.